Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add Utopia-V/mixagents --skill use-v4-flash-workergit clone --depth 1 https://github.com/Utopia-V/mixagentsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/utopia-v/mixagents/use-v4-flash-worker)<a href="https://agentmods.dev/skills/utopia-v/mixagents/use-v4-flash-worker"><img src="https://agentmods.dev/badge/skills/utopia-v/mixagents/use-v4-flash-worker.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00070 | $0.01003 |
| Opus 5 | $0.00035 | $0.00502 |
| Sonnet 5 | $0.00014 | $0.00201 |
| Haiku 4.5 | $0.00007 | $0.00100 |
Grade A, and why
use-v4-flash-worker scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 79 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Use V4 Flash Worker (legacy)
Check compatibility first
Read the active Codex version before staging an assignment or spawning a child.
If it is 0.149.0 or later, stop without calling a provider and direct the user
to mixagents-broker and its
installation guide.
If the version cannot be established, treat this native route as unqualified
and do not dispatch it.
Do not migrate the task to Broker automatically. The user must install and configure Broker before using its provider routes.
Select legacy work
- Select Flash only when the user requests it or its provider-declared 1M context changes whether a bounded, preferably read-only reading task is feasible. It is not the routine factual scout.
- Keep coupled reasoning, consequential decisions, verification, and final integration in the parent. Use a multimodal worker for images.
- The assignment crosses the configured DeepSeek data boundary. Sensitive material requires user authorization; credentials remain in the provider environment. Delegation never changes the parent's model or provider.
Deliver one self-contained job
- Build one complete assignment containing child identity, objective, scope, exclusions, available permissions, evidence or output contract, and stopping condition. Keep it in parent-owned execution state; do not publish it as user-visible commentary merely for transport.
- Pipe the assignment through stdin to the installed handoff script in
stagemode. Use the standard installed path below. If it is absent, inspect the effectiveSubagentStartHook matching^v4_flash_worker$and use the same reviewed script path with its mode changed fromhooktostage:- Windows:
powershell.exe -NoProfile -NonInteractive -ExecutionPolicy Bypass -File "<codex-home>\hooks\codex-deepseek-subagent\plaintext-handoff.ps1" -Mode stage - macOS/Linux:
python3 "<codex-home>/hooks/codex-deepseek-subagent/plaintext_handoff.py" --mode stage
- Windows:
- Require a successful stage result naming
v4_flash_worker. Treat a lock contender, an active pending or claimed item, quarantined state, or any other non-success result as a transport failure. Never spawn after a failed stage. Retry the complete stage only after the occupied state is explicitly clear, and spawn only after that new stage succeeds. - Immediately create the child through Codex's native
spawn_agentwith the exact agent typev4_flash_worker, a unique task name, andfork_turns="none". Do not replace this with a provider CLI, direct API call, or inherited root history. Keep all essential instructions in the staged assignment; let the spawn message only identify the trusted one-shot Hook. - Receive the child through Codex's native wait/callback path. Use one task-sized idle wait or callback; do not short-poll, duplicate the child work, or invent another return transport while it runs.
- Verify the returned contribution in proportion to the parent claim, then integrate it in the parent context.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago Changed · +7 lines · -4 tokens per session 681ed362139b
- 7d ago First seen · 72 lines · 74 tokens per session scan A d3b03622f8f8
use-v4-flash-worker is a skill published in the GitHub repository Utopia-V/mixagents (185 stars, last pushed 2d ago), licensed MIT. It adds 70 tokens to every session and 1,003 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
shortfilm-prompt
A five-part method for writing prompts for AI-generated videos, such as Sora or Seedance. It organizes each prompt into the subject, setting, visual style, camera movement, and timed shots with sound.
ao-workflow-runner
A workflow engine that reads YAML files and runs tasks assigned to different named roles. YAML is a text format for describing structured settings and steps.
subagent-driven-development
Execute plans via delegatetask subagents (2-stage review).
setup-dsh-bridge
Set up, repair, or reconfigure DSH Codex Bridge for the current Git project. Use when the user installs the plugin, asks to configure DSH, add or switch a model, change the default Profile, preview model configuration, roll back a Profile change, or when delegation reports missing/invalid Bridge configuration.
delegate-to-dsh
A workflow for sending a defined coding task to DeepSeek Harness, an alternate coding worker, in a separate worktree. Codex plans and reviews the returned patch and evidence.
Subagent Driven Development
Orchestrate specialized, autonomous AI agents to execute parallel subtasks with strict boundaries.